Question 1

Scaling Characteristics of Parallel Programs Ts is

Accepted Answer

Ts, or the serial portion of a parallel program, remains constant as it is the part of the program that cannot be parallelized.

Question 2

Speedup tends to saturate and efficiency _____ as a consequence of Amdahl's law.

Accepted Answer

Amdahl's law suggests that as more processors are added to a task with a fixed size, the speedup tends to saturate because only a portion of the task can be parallelized. The efficiency, which is the speedup per processor, decreases as the number of processors increases, due to the diminishing returns on adding more processors.

Question 3

Speedup obtained when the problem size is _______ linearly with the number of processing elements.

Accepted Answer

Speedup is typically obtained when the problem size increases linearly with the number of processing elements, as this allows for efficient parallelization and utilization of resources.

Question 4

The n × n matrix is partitioned among n processors, with each processor storing complete ___ of the matrix.

Accepted Answer

The answer of The n × n matrix is partitioned...

Question 5

cost-optimal parallel systems have an efficiency of ___

Accepted Answer

Cost-optimal parallel systems have an efficiency of 1, meaning they achieve the best possible performance without wasting resources.

Question 6

The n × n matrix is partitioned among n2 processors such that each processor owns a _____ element.

Accepted Answer

Each processor owns a single element when an n × n matrix is partitioned among n^2 processors, ensuring that each processor is responsible for exactly one element of the matrix.

Question 7

how many basic communication operations are used in matrix vector multiplication

Accepted Answer

The answer of how many basic communication operations are used...

Question 8

In DNS algorithm of matrix multiplication it used

Accepted Answer

The DNS (Dekel, Nassimi, Sahni) algorithm for matrix multiplication uses a 3D partition of the matrices involved. This approach allows for parallel computation by distributing the matrices across a three-dimensional processor grid, thus optimizing the multiplication process.

Question 9

In the Pipelined Execution, steps contain

Accepted Answer

Pipelined Execution steps contain normalization, communication, and elimination.

Question 10

the cost of the parallel algorithm is higher than the sequential run time by a factor of __

Accepted Answer

The answer of the cost of the parallel algorithm is...

Question 11

The load imbalance problem in Parallel Gaussian Elimination: can be alleviated by using a ____ mapping

Accepted Answer

Cyclic mapping helps in distributing the workload more evenly among processors, thus alleviating the load imbalance problem in Parallel Gaussian Elimination.

Question 12

A parallel algorithm is evaluated by its runtime in function of

Accepted Answer

Parallel algorithms are evaluated based on multiple factors including the input size, which determines the complexity of the problem; the number of processors, which affects how tasks are distributed and executed in parallel; and the communication parameters, which influence the efficiency of data exchange between processors. All these aspects are crucial for understanding and optimizing the runtime of parallel algorithms.

Question 13

For a problem consisting of W units of work, p__W processors can be used optimally.

Accepted Answer

For a problem consisting of W units of work, <= W processors can be used optimally because using more processors than the units of work does not improve performance due to the overhead of managing the extra processors and the lack of additional work for them to do.

Question 14

many interactions in oractical parallel programs occur in _____ pattern

Accepted Answer

Many interactions in practical parallel programs occur in a well-defined pattern, meaning that the interactions and communications between different parts of the program are structured and predictable, allowing for efficient parallel processing.

Question 15

efficient implementation of basic communication operation can improve

Accepted Answer

The answer of efficient implementation of basic communication operation can...

Question 16

efficient use of basic communication operations can reduce

Accepted Answer

Efficient use of basic communication operations can significantly reduce development effort by streamlining processes, improving team coordination, and minimizing misunderstandings, which can lead to faster and more effective development cycles.

Question 17

Group communication operations are built using_____ Messenging primitives.

Accepted Answer

Group communication operations in distributed systems or parallel computing environments are fundamentally built using point-to-point messaging primitives. These primitives allow for direct communication between two processes or nodes, serving as the basic building blocks for more complex group communication patterns.

Question 18

one processor has a piece of data and it need to send to everyone is

Accepted Answer

When one processor has a piece of data and it needs to send it to everyone, this is referred to as "one-to-all" broadcasting.

Question 19

the dual of one -to-all is

Accepted Answer

The dual of a one-to-all reduction is an all-to-one reduction, where multiple instances or problems are reduced to a single instance or problem.

Question 20

Data items must be combined piece-wise and the result made available at

Accepted Answer

Data items must be combined piece-wise and the result made available at the target processor finally. This is a fundamental concept in parallel computing where data is processed in parts across multiple processors and then combined at the target processor for the final output.

Quiz 4: Characteristics and Operations in Parallel Programming